EndoDx: A New Frontier in Endometriosis Diagnosis
I recently had the opportunity to speak with PhD student, Isabelle Chickanosky, to learn more about her passion project EndoDx and how CMI funding has helped accelerate the progress of the project.
Isabelle has always had a passion for research. Upon completing her bachelor’s degree from Carnegie Mellon University in Materials Science Engineering, she knew that she wanted to obtain her PhD. Due to her personal experiences with women’s health, she has a strong interest in endometriosis research. Endometriosis is the presence of tissue similar to that of the endometrium that spreads outside of the uterus forming lesions, adhesions, or endometriomas throughout the pelvis. She hoped that this type of research could be pursued during her PhD experience.
Upon applying to the University of Pittsburgh PhD in Bioengineering program, Isabelle was interviewed by Professor David Vorp about joining his lab at the Swanson School of Engineering. During the interview, Dr. Vorp wanted to learn more about Isabelle’s research goals. Through this conversation, Isabelle shared her strong interest in women’s health research, especially about endometriosis. Even though Dr. Vorp’s lab did not do endometriosis research, Isabelle was interested in working in his lab. “I decided that this experience would expand my toolbox as a researcher. I was particularly interested in the machine learning that was going on in the lab. This was something I wanted to learn more about.”
About four months after Isabelle began her studies at the University of Pittsburgh, Dr. Vorp approached her after a meeting. He explained that his wife was scheduled to have endometriosis diagnostic surgery after living with symptoms for about five years. After having her symptoms dismissed as something else for so long, she was hopeful that the surgery would give her an answer. Since Dr. Vorp knew of Isabelle’s interest in women’s health research and personal experience with endometriosis, he expressed interest in wanting to learn as much as possible about endometriosis and how to improve patient outcomes.
The personal connections that both Isabelle and Dr. Vorp have to endometriosis led to conversations of, “what do we know and how can we add to the field?”. They began having meetings once a week to understand the endometriosis field. Currently, the protocol for endometriosis is an invasive procedure to obtain a sample of tissue that can then determine the disease state. They wanted to investigate possible non-invasive applications of machine learning to diagnose endometriosis. They began setting up connections within the Pittsburgh area, starting with Isabelle’s surgeon, Dr. Nicole Donnellan at Magee Women’s Hospital who has since become a main collaborator on the project.
Through these conversations and connections, the idea of a new company, EndoDx, was born. The project has been a collaborative effort between Dr. David Vorp (PI), Dr. Donellan (clinical collaborator), Dr. John Harris (clinical and genomics collaborator), and Isabelle Chickanosky (PhD student). EndoDx is developing a machine learning diagnosis tool for endometriosis. The machine learning algorithm works by taking in patient specific inputs such as biomarkers, clinical data, and survey data to predict if a patient has endometriosis. If it determines that a patient has endometriosis, it can then segment how severe it is. Endometriosis is divided into four stages based on the number of lesions and depth of infiltration with Stage I being minimal and Stage IV being severe. This is an important aspect of the algorithm as the stage state does not always correspond to patient symptoms. By using machine learning algorithms, the dots can be connected between the different inputs so that better patient outcomes can be achieved.
EndoDx was granted $15,000 from the Center for Medical Innovation (CMI) in 2022. With the funds granted, the team added 115 patients with pelvic pain or infertility issues. So far, the model has about 70% success rate in accurately predicting endometriosis diagnosis. Additionally, it can categorize patients who have endometriosis into two groups regarding disease stage (stage 1 and 2 or stage 3 and 4) at about 85% accuracy. These success rates are a result of implementing clinical and survey data into the model patients can provide themselves. By including the biomarker data in the AI model, Isabelle believes it will further improve the predictive value.
EndoDx has also received other funding to accelerate the progress of the project. It received $15,000 from the Michael G. Wells Student Healthcare Competition and $50,000 from the Virginal Kaufman PAIN Research Challenge. All these funding sources have been instrumental in getting the project off the ground and expanding patient enrollment to better train the machine learning model. The group recently applied for a R21 NIH grant which will allow for the enrollment of even more patients and study more biomarkers. The first publication of the EndoDx project is also underway.
Isabelle is extremely hopeful that EndoDx will have a huge impact one day. “Being at the University of Pittsburgh, I have a unique opportunity to spin this work out into a start-up company by developing a clinically implementable tool.” She envisions EndoDx being used in gynecologist offices not just specialized endometriosis clinics. Isabelle went on to say, “Maybe one day it could be an at home test – similar to taking an at home pregnancy test.” Overall, the goal is that one day EndoDx would reach people globally, providing a new diagnostic tool and reducing the number of invasive surgeries for women who do not have endometriosis despite similar symptoms.